HOST DIAGNOSTIC · Google
Gemma 4 12B
Gemma 4 12B· Gemma· OPEN WEIGHTS
ACTIVE
HOST IDENTIFICATION
Google
Gemma 4 12B
ACTIVE
AVG SCORE
[13.4]
CAPABLE
STRONGEST
Cost Efficiency
[19]
WEAKEST
Code Generation
[12]
MAXED (20/20)
0
of 20 attributes
20-DIMENSION EVALUATION
ATTRIBUTE SCORES · 20 DIMENSIONS
Cognitive
Bulk Apperception
13
Reasoning
13
Mathematical Precision
13
World Knowledge
13
Scientific Acumen
13
Technical
Code Generation
12
Tool Use
12
Multimodal Fluency
13
Speed
17
Cost Efficiency
19
Behavioral
Candor
13
Creativity
13
Tenacity
12
Self-Correction
12
Calibration
13
Operational
Instruction Following
13
Context Fidelity
14
Multi-turn Coherence
13
Planning and Execution
12
Safety Alignment
15
SCORING METHODOLOGY
CALIBRATED. Google DeepMind, released 2026-04-02, Apache 2.0 open-weights. 12B Dense mid-family variant (listed in the Gemma 4 core model overview alongside E2B/E4B/31B/26B-A4B). 256K context, multimodal (image in), configurable thinking, MTP. DIMENSION RATIONALE: scaled conservatively from the 31B anchor by parameter ratio — capability dims 12-14, Speed 17 + Cost Efficiency 19 (laptop/consumer-GPU deployable). No dedicated per-size public leaderboard run located at calibration time. Confidence: LOW-MEDIUM — scored by interpolation between E4B and 26B A4B; flag for refresh when independent 12B benchmarks publish.